Arnab Bose

Spoke at Code with @claudeai about how @asana's AI Teammates are built on managed agents. The split: we bring the enterprise context, multiplayer UX, and security controls for shared memory. Claude brings the reasoning and the muscle to actually finish complex work.

Scheduled opposite Boris and Jared on the main stage at Code with Claude today. Come to my breakout anyway? I'll be showing what we've built with Asana AI Teammates 🙏 - I promise it will be 🤯

I recently joined the  @awscloud (AWS) for Software Companies podcast to talk about what it takes for AI to do work in a way teams can actually trust. Most companies are still trying to bolt agents onto fragmented systems. Those agents can generate output, but they lack the shared context, checkpoints, and controls needed to operate across a business. That’s a big part of why I’m excited about what we’re building at @asana. Asana gives AI Teammates organizational context — goals, projects, tasks, dependencies, and owners — so they can do more than respond to a prompt. They can understand how work actually gets done. And because AI Teammates are multiplayer by design, that context doesn’t live in a silo. Teams can see what agents are doing, guide them, correct them, and improve outcomes together. Over time, the intelligence stays in the system and compounds across the team. That’s why our partnership with Amazon Web Services (AWS) matters. AWS and Asana each bring a critical piece of the puzzle: world-class infrastructure on one side, and the organizational context that AI needs to provide real value on the other. Together, that’s how we help move enterprise AI from experimentation to execution. open.spotify.com/episode/6gWYVS…

@tanayj @brackin asana.com/resources/ai-t…

@brackin surprised asana, things, etc haven't invested more in this x.com/tanayj/status/…

I need a to-do list where I can assign tasks to agents (town, claude, etc), and agents can assign tasks to me

@PaulYacoubian @therealredman So sorry to hear this Paul.

@asana built AI Teammates on Managed Agents: agents that work inside Asana as part of your team, picking up assigned tasks. Offloading the infrastructure let them put engineering time into the multiplayer experience.

Here is a link to the Q&A with Anthropic: claude.com/customers/asan…

I had a great conversation with the @AnthropicAI team. The power of a frontier model like Opus 4.6 is immense but to harness it so businesses achieve outcomes requires context, checkpoints and controls. Asana AI Teammates shift teams from individual prompts to super-charged execution. Here’s how: - Multiplayer by design. They live in the project, not a private chat. When one person shifts a priority, the AI Teammate updates the roadmap for the whole team. - Zero coordination overhead. By absorbing the "coordination tax"—the chasing, status pings, and manual handoffs—AI Teammates drive higher throughput with less rework and manual orchestration. - Institutional memory. Knowledge stays in the system. As your team gives feedback, the AI Teammate gets smarter for everyone, preserving context as your organization scales. - Governed execution. Innovation shouldn't create risk. These agents operate within your existing permissions and security rules—no new frameworks or guardrails to manage.

@asana was just named #1 in the Workplace category on @FastCompany 's Most Innovative Companies list for 2026. We made a bet that AI agents shouldn't just be personal copilots. They should be shared teammates embedded in how real teams actually work. That meant building on top of the Work Graph, giving AI the full context of not just who is doing what but also the blueprint for how work is executed across the business.

Read more about our approach here: asana.com/inside-asana/a…

While AI models are capable, they often fail in professional settings due to an "accountability gap." @asana aims to bridge this by integrating AI into its existing "work graph" architecture. Key Pillars of Asana's AI Strategy Shared Context: Unlike tools that only see what a single user sees, AI Teammates live within projects and tasks. They build "working knowledge" by observing interactions between multiple human collaborators. Transparency & Accountability: Every action taken by an AI Teammate is recorded in an action log. They follow the same permission models as humans and require explicit approval for sensitive actions. Human-Like Collaboration: Teammates are treated as members of the org; they are assigned tasks, write comments, and appear in activity feeds. Simple, Inspectable Memory: Instead of complex black-box databases, Asana uses a text-based memory system that users can view and edit directly on the Teammate's profile to correct or refine its behavior. Nondeterminism: Rather than forcing teams into rigid, pre-set flowcharts, the AI learns a specific team’s unique way of working (e.g., how they use subtasks vs. sections) to provide tailored assistance. The Bottom Line Asana believes the most successful AI agents won't just have the best "demos," but will be the ones that mirror human teamwork: structured, visible, and accountable.

If you'd like to learn more about what we launched, you can read the full announcement here: asana.com/resources/ai-t…

Yesterday we announced the general availability of @Asana AI Teammates. These aren’t copilots. They’re AI agents that operate directly inside your team’s workflows — with the context and controls needed to actually execute work. Quick overview 👇

An engineer at Anthropic wrote a spec, pointed Claude at an Asana board, and went home. Claude broke the spec into tickets, spawned agents for each one, and they started building independently. When the agent is confused it runs git-blame and messages the right engineers in Slack. By Monday the agents finished the plugin feature. That's one example of how the best engineers are shipping software right now. Developers will soon orchestrate 50 AI agents in parallel and the difference between a good engineer & a great one would come down to specs. You can't write a spec that holds up at that scale without genuinely understanding what you're building at a deeper level. The next-gen developer who understands the fundamentals, can architect well and orchestrate agent is going to be a 1000x developer!

Give an agent real context—shared workflows, inherited permissions, visible checkpoints—and trust follows naturally. Read more @VerdictUK verdict.co.uk/trusting-ai-ag…

That's why only 6% of companies fully trust AI agents today. Not because the technology isn't powerful enough. Because most agents operate in a vacuum—disconnected from the systems, the people, and the history that actually matter.

The capability of AI agents isn't the problem. Context is. You can't LLM your way past 10 years of customer data. And without that deep context—the workflows, the permissions, the institutional knowledge—you don't get trust.

@spenserskates @asana Thanks for joining our Company Kickoff! The team loved our conversation and your candid insights!

Just did a fireside chat with @ArBose at @asana CKO. Big future for the company that cracks multiplayer agents first

Read more about our approach to making agents multiplayer here: venturebeat.com/orchestration/…

Asana launched Asana AI Teammates last year with the philosophy that, just like humans, AI agents should be plugged directly into a team, have enterprise context and shared knowledge and memory to be truly successful.